Fix sync-and-exit option logic (#628)

* fix sync-and-exit logic

* fix reading from doneSync channel

* clean up

* make statusd before using it

* move syncAndStopNode to different file

* change log level in travis chain sync command

* do not use fmt but log

* add progress log and set timeout to 20 mins

* set datadir to .ethereumtest

* fix datadir

.travis.yml

@@ -27,7 +27,9 @@ jobs:
     # to still run, but currently does not work due to a bug
     if: type != pull_request
     script:
-      - ./build/bin/statusd -les -networkid 4 -sync-and-exit
+      # sync the chain first; it will time out after 20 minutes; Rinkeby is networkid=4
+      - make statusgo
+      - ./build/bin/statusd -datadir .ethereumtest/Rinkeby -les -networkid=4 -sync-and-exit=20 -log WARN -standalone=false -discovery=false
       - make test-e2e networkid=4
 cache:
   directories:

cmd/statusd/main.go

@@ -114,17 +114,25 @@ func main() {
 
 	// Sync blockchain and stop.
 	if *syncAndExit >= 0 {
-		exitCode := syncAndStopNode(backend.NodeManager(), *syncAndExit)
+		exitCode := syncAndStopNode(interruptCh, backend.NodeManager(), *syncAndExit)
+		// Call was interrupted. Wait for graceful shutdown.
+		if exitCode == -1 {
+			if node, err := backend.NodeManager().Node(); err == nil && node != nil {
+				node.Wait()
+			}
+			return
+		}
+		// Otherwise, exit immediately with a returned exit code.
 		os.Exit(exitCode)
 	}
 
-	// wait till node has been stopped
 	node, err := backend.NodeManager().Node()
 	if err != nil {
 		log.Fatalf("Getting node failed: %v", err)
 		return
 	}
 
+	// wait till node has been stopped
 	node.Wait()
 }
 
@@ -170,34 +178,6 @@ func startCollectingStats(interruptCh <-chan struct{}, nodeManager common.NodeMa
 	}
 }
 
-func syncAndStopNode(nodeManager common.NodeManager, timeout int) (exitCode int) {
-	log.Println("Node will synchronize and exit")
-	if timeout < 0 {
-		log.Println("Sync and stop error: negative timeout value")
-		return 1
-	}
-	var err error
-	if timeout == 0 {
-		err = nodeManager.EnsureSync(context.Background())
-	} else {
-		ctx, cancel := context.WithTimeout(context.Background(), (time.Duration)(timeout)*time.Minute)
-		err = nodeManager.EnsureSync(ctx)
-		defer cancel()
-	}
-	if err != nil {
-		log.Printf("Sync error: %v", err)
-		exitCode = 1
-	}
-	var done <-chan struct{}
-	done, err = nodeManager.StopNode()
-	if err != nil {
-		log.Printf("Stop node err: %v", err)
-		return 1
-	}
-	<-done
-	return
-}
-
 // makeNodeConfig parses incoming CLI options and returns node configuration object
 func makeNodeConfig() (*params.NodeConfig, error) {
 	devMode := !*prodMode

cmd/statusd/sync.go

@@ -0,0 +1,57 @@
+package main
+
+import (
+	"context"
+	"log"
+	"time"
+
+	"github.com/status-im/status-go/geth/common"
+)
+
+func createContextFromTimeout(timeout int) (context.Context, context.CancelFunc) {
+	if timeout == 0 {
+		return context.WithCancel(context.Background())
+	}
+
+	return context.WithTimeout(context.Background(), time.Duration(timeout)*time.Minute)
+}
+
+// syncAndStopNode tries to sync the blockchain and stop the node.
+// It returns an exit code (`0` if successful or `1` in case of error)
+// that can be used in `os.Exit` to exit immediately when the function returns.
+// The special exit code `-1` is used if execution was interrupted.
+func syncAndStopNode(interruptCh <-chan struct{}, nodeManager common.NodeManager, timeout int) (exitCode int) {
+	log.Printf("syncAndStopNode: node will synchronize the chain and exit (timeout %d mins)", timeout)
+
+	ctx, cancel := createContextFromTimeout(timeout)
+	defer cancel()
+
+	doneSync := make(chan struct{})
+	errSync := make(chan error)
+	go func() {
+		if err := nodeManager.EnsureSync(ctx); err != nil {
+			errSync <- err
+		}
+		close(doneSync)
+	}()
+
+	select {
+	case err := <-errSync:
+		log.Printf("syncAndStopNode: failed to sync the chain: %v", err)
+		exitCode = 1
+	case <-doneSync:
+	case <-interruptCh:
+		// cancel context and return immediately if interrupted
+		// `-1` is used as a special exit code to denote interruption
+		return -1
+	}
+
+	done, err := nodeManager.StopNode()
+	if err != nil {
+		log.Printf("syncAndStopNode: failed to stop the node: %v", err)
+		return 1
+	}
+	<-done
+
+	return
+}

geth/node/manager.go

@@ -532,25 +532,33 @@ func (m *NodeManager) EnsureSync(ctx context.Context) error {
 	if m.config.NetworkID == params.StatusChainNetworkID {
 		return nil
 	}
-	if m.lesService == nil {
-		return errors.New("LightEthereumService is nil")
-	}
+
 	return m.ensureSync(ctx)
 }
 
 func (m *NodeManager) ensureSync(ctx context.Context) error {
-	downloader := m.lesService.Downloader()
+	les, err := m.LightEthereumService()
+	if err != nil {
+		return fmt.Errorf("failed to get LES service: %v", err)
+	}
+
+	downloader := les.Downloader()
 	if downloader == nil {
 		return errors.New("LightEthereumService downloader is nil")
 	}
+
 	progress := downloader.Progress()
-	if progress.CurrentBlock >= progress.HighestBlock {
-		log.Debug("Synchronization completed")
+	if m.PeerCount() > 0 && progress.CurrentBlock >= progress.HighestBlock {
+		log.Debug("Synchronization completed", "current block", progress.CurrentBlock, "highest block", progress.HighestBlock)
 		return nil
 	}
 
 	ticker := time.NewTicker(tickerResolution)
 	defer ticker.Stop()
+
+	progressTicker := time.NewTicker(time.Minute)
+	defer progressTicker.Stop()
+
 	for {
 		select {
 		case <-ctx.Done():
@@ -566,13 +574,13 @@ func (m *NodeManager) ensureSync(ctx context.Context) error {
 			}
 			progress = downloader.Progress()
 			if progress.CurrentBlock >= progress.HighestBlock {
-				log.Debug("Synchronization completed")
+				log.Info("Synchronization completed", "current block", progress.CurrentBlock, "highest block", progress.HighestBlock)
 				return nil
 			}
-			log.Debug(
-				fmt.Sprintf("Synchronization is not finished yet: current block %d < highest block %d",
-					progress.CurrentBlock, progress.HighestBlock),
-			)
+			log.Debug("Synchronization is not finished", "current", progress.CurrentBlock, "highest", progress.HighestBlock)
+		case <-progressTicker.C:
+			progress = downloader.Progress()
+			log.Warn("Synchronization is not finished", "current", progress.CurrentBlock, "highest", progress.HighestBlock)
 		}
 	}
 }